[OpenWrt-Devel] [PATCH] mvebu: add support for GL.iNet GL-MV1000
Tomasz Maciej Nowak
tomek_n at o2.pl
Mon Apr 13 16:14:21 EDT 2020
Hi Li Zhang.
This version looks much better than the first one,
some comments inline.
W dniu 13.04.2020 o 12:26, Li Zhang pisze:
> This patch adds supports for GL-MV1000.
>
> Specification:
> - SOC: Marvell Armada 88F3720 (1GHz)
> - Flash: 16MB (W25Q128FWSIG)
> - RAM: 1GB DDR4
> - Ethernet: 3x GE (1 WAN + 2 LAN)
> - EMMC: 8GB EMMC (KLM8G1GETF-B041)
> - MicroSD: 1x microSD slot
> - USB: 1x USB 2.0 port(TypeA),1x USB 3.0 port(TypeC)
> - Button: 1x reset button,1x slide switch
> - LED: 3x greed LED
> - UART: 1x UART on PCB (JP1: 3.3V, RX, TX, GND)
>
> Signed-off-by: Li Zhang <li.zhang at gl-inet.com>
> ---
> package/boot/uboot-envtools/files/mvebu | 3 +
> .../cortexa53/base-files/etc/board.d/02_network | 3 +-
> .../boot/dts/marvell/armada-gl-mv1000-emmc.dts | 68 ++++++++++++++++++++++
> target/linux/mvebu/image/Makefile | 9 +++
> target/linux/mvebu/image/cortexa53.mk | 9 +++
> target/linux/mvebu/image/gen_mvebu_sdcard_img.sh | 6 ++
> .../mvebu/image/generic-arm64-emmc.bootscript | 12 ++++
> 7 files changed, 109 insertions(+), 1 deletion(-)
> create mode 100644 target/linux/mvebu/files-4.19/arch/arm64/boot/dts/marvell/armada-gl-mv1000-emmc.dts
> create mode 100644 target/linux/mvebu/image/generic-arm64-emmc.bootscript
>
> diff --git a/package/boot/uboot-envtools/files/mvebu b/package/boot/uboot-envtools/files/mvebu
> index 7902384..9d23c77 100644
> --- a/package/boot/uboot-envtools/files/mvebu
> +++ b/package/boot/uboot-envtools/files/mvebu
> @@ -24,6 +24,9 @@ globalscale,espressobin-v7-emmc|\
> marvell,armada8040-mcbin)
> ubootenv_add_uci_config "/dev/mtd0" "0x3f0000" "0x10000" "0x10000" "1"
> ;;
> +glinet,gl-mv1000-emmc)
> + ubootenv_add_uci_config "/dev/mtd1" "0x0" "0x8000" "0x8000" "1"
> + ;;
> linksys,caiman|\
> linksys,cobra|\
> linksys,shelby)
> diff --git a/target/linux/mvebu/cortexa53/base-files/etc/board.d/02_network b/target/linux/mvebu/cortexa53/base-files/etc/board.d/02_network
> index ba4b930..e5ff667 100755
> --- a/target/linux/mvebu/cortexa53/base-files/etc/board.d/02_network
> +++ b/target/linux/mvebu/cortexa53/base-files/etc/board.d/02_network
> @@ -14,7 +14,8 @@ case "$board" in
> globalscale,espressobin|\
> globalscale,espressobin-emmc|\
> globalscale,espressobin-v7|\
> -globalscale,espressobin-v7-emmc)
> +globalscale,espressobin-v7-emmc|\
> +glinet,gl-mv1000-emmc)
> ucidef_set_interfaces_lan_wan "lan0 lan1" "wan"
> ;;
> marvell,armada-3720-db|\
> diff --git a/target/linux/mvebu/files-4.19/arch/arm64/boot/dts/marvell/armada-gl-mv1000-emmc.dts b/target/linux/mvebu/files-4.19/arch/arm64/boot/dts/marvell/armada-gl-mv1000-emmc.dts
> new file mode 100644
> index 0000000..fe01dfe
> --- /dev/null
> +++ b/target/linux/mvebu/files-4.19/arch/arm64/boot/dts/marvell/armada-gl-mv1000-emmc.dts
> @@ -0,0 +1,68 @@
Please add a license in SPDX format, a common one is dual MIT and
GPL-2.0+.
> +/*
> + * Device Tree file for GL.iNet GL-MV1000
> + */
> +
> +#include "armada-3720-espressobin.dts"
This is different device from ESPRESSObin altogether, please make it
stand-alone dts (copy espressobin dts, add required nodes and
remove/disable nodes for devices not present on the GL-MV1000).
Would be good to add all LEDs and buttons as nodes, which seem to be
GPIO controlled as in this picture:
https://static.gl-inet.com/docs/en/3/hardware/mv1000/mv1000.png
but that's not mandatory.
> +
> +/ {
> + model = "GL.inet GL-MV1000 (Marvell)";
Instead of Marvell more apropriate would be Brume, since this is used
on Your website.
> + compatible = "glinet,gl-mv1000-emmc";
Add here to compatible array in second place "marvell,armada3720".
Are there any boards in cutomers hands which do not have SD card slot
or eMMC soldered? There is no point in indicating the eMMC presence
when all sold boards are the same. In case of ESPRESSObin it's there
because various hardware versions. So please remove all -emmc suffixes
from files and file names across this patch if that's the case. One
exeption would be if U-Boot searches for exctly this name
(armada-gl-mv1000-emmc.dts) when booting, but looking at the usage of
boot.scr that's not the case.
> +};
> +
> +&spi0 {
> + status = "okay";
> +
> + flash at 0 {
> + reg = <0>;
> + compatible = "jedec,spi-nor";
> + spi-max-frequency = <104000000>;
> + m25p,fast-read;
> + partitions {
> + compatible = "fixed-partitions";
> + #address-cells = <1>;
> + #size-cells = <1>;
> +
> + partition at 0 {
> + label = "u-boot";
> + reg = <0 0xf0000>;
> + };
> +
> + partition at f0000 {
> + label = "u-boot-env";
> + reg = <0Xf0000 0x8000>;
> + };
> +
> + art: partition at f8000 {
> + label = "art";
The name of this partition is rather strange, since the hardware
specification doesn't mention any PCIe or SDIO connected wireless
cards (don't know if there's any USB card without eeprom). Is
there any calibration data stored or only MAC address/addresses?
Does U-Boot also read information from this partition? Anyway
that's not a remark to change that name, I'm just curious.
> + reg = <0xf8000 0x8000>;
> + };
On previous patch, there were additionnal partitions in dts
without -emmc suffix : dtb, firmware. If that space would be
unused it's better to add them here, so later You could
introduce images that could be written in SPI flash.
> +
> + };
> + };
> +};
> +
> +&sdhci1 {
> + wp-inverted;
> + bus-width = <4>;
> + cd-gpios = <&gpionb 17 GPIO_ACTIVE_LOW>;
> + marvell,pad-type = "sd";
> + no-1-8-v;
> + vqmmc-supply = <&vcc_sd_reg1>;
> + status = "okay";
> +};
> +
> +
> +&sdhci0 {
> + bus-width = <8>;
> + mmc-ddr-1_8v;
> + mmc-hs400-1_8v;
> + non-removable;
> + no-sd;
> + no-sdio;
> + marvell,pad-type = "fixed-1-8v";
> + status = "okay";
> +};
> +
> +ð0 {
> + mtd-mac-address = <&art 0x0>;
> +};
> diff --git a/target/linux/mvebu/image/Makefile b/target/linux/mvebu/image/Makefile
> index ef92748..b848049 100644
> --- a/target/linux/mvebu/image/Makefile
> +++ b/target/linux/mvebu/image/Makefile
> @@ -107,6 +107,15 @@ define Device/Default-arm64
> KERNEL := kernel-bin
> endef
>
> +define Device/Default-arm64-emmc
Please don't add generic difinition for single device where only
image name and recipe are slightly changed, please remove it.
> + BOOT_SCRIPT := generic-arm64-emmc
> + DEVICE_DTS_DIR := $(DTS_DIR)/marvell
> + IMAGES := emmc.img
> + IMAGE/emmc.img := boot-scr | boot-img-ext4 | sdcard-img-ext4 | append-metadata
> + KERNEL_NAME := Image
> + KERNEL := kernel-bin
> +endef
> +
> define Device/NAND-128K
> BLOCKSIZE := 128k
> PAGESIZE := 2048
> diff --git a/target/linux/mvebu/image/cortexa53.mk b/target/linux/mvebu/image/cortexa53.mk
> index 77f5c79..4dfd0b4 100644
> --- a/target/linux/mvebu/image/cortexa53.mk
> +++ b/target/linux/mvebu/image/cortexa53.mk
> @@ -69,3 +69,12 @@ define Device/methode_udpu
> BOOT_SCRIPT := udpu
> endef
> TARGET_DEVICES += methode_udpu
> +
> +define Device/glinet_gl-mv1000-emmc
> + $(call Device/Default-arm64-emmc)
Please call "Default-arm64" here. You can override image recipe
here adding (don't know if that'll be necessary):
BOOT_SCRIPT := generic-arm64-emmc
IMAGES := emmc.img
IMAGE/emmc.img := boot-scr | boot-img-ext4 | sdcard-img-ext4 | append-metadata
> + DEVICE_TITLE := GL.iNet GL-MV1000 EMMC
> + DEVICE_DTS := armada-gl-mv1000-emmc
> + SUPPORTED_DEVICES := glinet,gl-mv1000-emmc
> +endef
> +TARGET_DEVICES += glinet_gl-mv1000-emmc
> +
> diff --git a/target/linux/mvebu/image/gen_mvebu_sdcard_img.sh b/target/linux/mvebu/image/gen_mvebu_sdcard_img.sh
> index 842e591..50028fe 100755
> --- a/target/linux/mvebu/image/gen_mvebu_sdcard_img.sh
> +++ b/target/linux/mvebu/image/gen_mvebu_sdcard_img.sh
> @@ -51,6 +51,12 @@ while [ "$#" -ge 3 ]; do
> shift; shift; shift
> done
>
> +model=''
> +model=$(echo $OUTFILE | grep "gl-mv1000-emmc")
> +[ "$model" != "" ] && {
> + ptgen_args="$ptgen_args -t 83 -p 7093504"
Please don't add device hacks here it's ment to be as generic
as posible. Instead, add possibility to override CONFIG_TARGET_ROOTFS_PARTSIZE
as parameter to "boot-img-ext4" command.
> +}
> +
> head=16
> sect=63
>
> diff --git a/target/linux/mvebu/image/generic-arm64-emmc.bootscript b/target/linux/mvebu/image/generic-arm64-emmc.bootscript
> new file mode 100644
> index 0000000..4de4d39
> --- /dev/null
> +++ b/target/linux/mvebu/image/generic-arm64-emmc.bootscript
> @@ -0,0 +1,12 @@
> +setenv bootargs "root=/dev/mmcblk0p2 rw rootwait"
> +
> +if test -n "${console}"; then
> + setenv bootargs "${bootargs} ${console}"
> +fi
> +
> +setenv mmcdev 0
Does the U-Boot only load "boot.scr" from first mmcdev
(I assume that's eMMC) or also searches for it on SD card? If
searched on both devices, there's no need to add this file, use
generic-arm64.bootscript. That way image will be applicable to
both storage mediums, just write this information in commit
message to inform users. If U-Boot searches for "boot.scr" only
on eMMC, then rename this file to gl-mv1000.bootscript and set
it in BOOT_SCRIPT variable.
> +
> +load mmc ${mmcdev}:1 ${fdt_addr} @DTB at .dtb
> +load mmc ${mmcdev}:1 ${kernel_addr} Image
> +
> +booti ${kernel_addr} - ${fdt_addr}
>
Regards, Tomasz
--
TMN
_______________________________________________
openwrt-devel mailing list
openwrt-devel at lists.openwrt.org
https://lists.openwrt.org/mailman/listinfo/openwrt-devel
More information about the openwrt-devel
mailing list